FIAT/IFTA World Conference 2022

Archive data offers a wealth of information that can be used to answer a wide range of questions. For example, to analyse the guest lists of TV programmes to see how the representation of demographic groups changes over time. Or to count the occurrences of locations in wartime newspapers to investigate the geopolitical balance in the reports. This type of quantitative analysis can be performed on a much larger number of archive items than a researcher could reasonably analyse by hand.

AI goes a step further, in not only analysing and aggregating existing data, but learning to create new metadata, and even draw inferences from it. For instance, AI can learn to recognise faces to see who appears in a video, or to classify news articles into categories. A well trained AI can even detect the sentiment expressed in a piece of text.

With these techniques, it may seem that any possible question can be answered. However, there are important limitations. For example, it is easy to search for the term ‘European Union’ in text, but hard to discover how the concept of a union of European countries evolved, when this may have been described in many different ways. It is challenging to distinguish if the sentence ‘That’s a really good idea’ was said sarcastically. Finally, an important question is how representative an analysis of archive material is, given that it is a curated selection.

At Sound and Vision, we are developing support for researchers who aim to answer research questions with quantitative data. In this presentation, we will share our experiences of the questions we can answer, and the ones we can’t, and why. We will also discuss how we are investigating the concept of data stories that combine the traditional qualitative research methods with quantitative analysis to get the best out of both.